/ Social / Machine Learning /


Enable someone to accurately predict which photo will be the most popular out of a series of shots

Attention Measurement

  1. Scan all previous photo every posted by you.
  2. Build algorithm that measures surprise tailored to you and your way of sorting information.
  3. Recommend the photo that generates the most novelty from all the others.

Check the Quality first


Use a machine learning algorithm to look for:
1 – evidence of a bad quality photo, such as: blur , low resolution, one colored (completely white or black).
2 – human faces, their expresssions, conflicting expressions, and in particular Surprised faces.
3 – animals, like cats for example, and the presence (or not ) of Humans in the same picture
4 – Famous or widely know persons (needs internet)

And expose the algorithm to already popular images, so it can relate these features between each other, and do it autonomously.
OpenCV and TensorFlow might work well with this.



With Mobile app – Pass , you can pass or recommend few of your selected photographs or may be anything to your friends. Then each friend in turn passes them to someone else. The reach of any photo(or link) can therefore be judged by how many times a user recommends that photograph(media or link) to his or her friends.
Each user will get points to see and pass on any link or photograph they see. These points may be used to offer them any discounts to encourage them to see more. To reduce the clutter of links and photographs that appears in the list of any user to pass on , there may be some kind of limits on how many items you can pass per hour.

Machine Learning techniques.


There are two ways to do this:
1. Using regression model to predict the popularity as a significant number.

2. To classify the image as a popular, moderate popular or least popular using classification models.

For both of the models what we need is some features of the images like quality, dimension, website it’s posted on, type, etc.
and at the same time a label to all those features as of a significant number to show popularity or a labeled class to classify.

Progressive List


You could make a list of the features that the photo have to comply with to become popular or not popular, and update it as the Trends changes.

[I’m Italian…]


Waseem Raja

Professional photographer group working together by reviewing the picture of the customer . Where customer pay them to review it to know which picture will be more popular of all the picture he or she got . No one will see it except them because he will be uploading it in instagram later . Photographer or some team helps them by providing tagname which will be popularly used for this type of pictures . Which searching in instagram with those tagname . He’s image will get more likes and popularity for him . Simple rule if he want to be popular than he need to pay for expert reviews . Unpaid customer can also use it but there review will take more time .



If a series was sent immediately to fellow users and preliminary vote was completed within seconds, this could provide good insight into which photo would be the most popular. If you want to use the platform, you also need to vote on images in order to be able to submit them.

Cloud Gram


It’s early days, but Google Cloud Vision can extract a pretty good amount of information from an image. If we could process a significant amount of top images (highest ‘likes to followers’ ratio to avoid just getting famous people’s images) we could get a baseline range of meta data around emotion, colour, objects, etc. I’m not sure how that will then translate to selecting the most popular from a series of images, but it should highlight the image that is ‘most like’ the images that have the highest ‘like to followers’ ratio.

Submit your own idea!

Submit your idea to earn a future revenue share!

+ Submit Idea